This paper proposes a new, more robust, experiment to test for the presence of hyperbolic discounting. Recently, a growing literature has studied intertemporal choice when individuals discount the future hyperbolically. These preferences generate dynamically inconsistent choices, in contrast with the usual assumption of exponential discounting, where this issue cannot arise. Hyperbolic discounting is justified based on experimental evidence of individual self-control problems. We argue that this interpretation depends crucially on the absence of uncertainty. We show that, once uncertainty is included, the observed behavior is compatible with exponential discounting. We then test for the presence of hyperbolic discounting in a new experiment that controls for uncertainty. The experiment offers two choice sets, the second being a strict subset of the first. Exponential discounters will (possibly weakly) prefer the largest one. Hyperbolic discounters, in contrast, will (strictly) prefer the second set because its design makes it equivalent to a commitment technology. The experiment is conducted on a sample of undergraduate students. Our results suggest that hyperbolic behavior is more difficult to find than implied by previous experiments.